Clinical synthesis based on the landmark review Millar 2022 Nature Reviews, the UK FROST trial 2020 Lancet, the network meta-analysis Challoumas 2020 JAMA Network Open and the 2023-2026 data on the diabetic subgroup.

Frozen shoulder, known in English as frozen shoulder or adhesive capsulitis, is a painful and disabling condition characterised by a progressive, substantial loss of active and passive movement at the glenohumeral joint. Recognised since 1934 (Codman), it paradoxically remains poorly understood in aetiopathogenic terms, as the landmark review Millar 2022, published in Nature Reviews Disease Primers, points out.¹

How is this condition defined, who does it affect and what are the risk factors?

Frozen shoulder is defined as an inflammatory fibrosis of the glenohumeral joint capsule, causing thickening and retraction of the capsule and a painful restriction of movement.^1,2 Its prevalence in the general population is estimated at between 2 % and 5 %, affecting women preferentially (OR about 1.6) between 40 and 60 years.¹ Idiopathic forms are the most frequent, but secondary forms occur after trauma, shoulder surgery (particularly of the cuff) or prolonged immobilisation.³

Several systemic and local risk factors are clearly established. The most powerful and best documented is diabetes mellitus (T1DM and T2DM), with a prevalence reaching 10 to 38 % across series, and forms that are consistently more severe and more resistant to treatment.^4,5 Thyroid disorders are the second axis: the Chuang 2023 meta-analysis (10 case-control studies, 127,967 patients, JSES) finds more thyroid disease in patients with frozen shoulder (OR 1.87; 95 % CI 1.37-2.57), significantly so for hypothyroidism (OR 1.92) and subclinical hypothyroidism (OR 2.56), but not for hyperthyroidism (OR 1.42; 95 % CI 0.63-3.22).⁶

Other important risk factors include Dupuytren's disease (a robust association suggesting a genetic predisposition to fibrosis), prolonged immobilisation after a fracture or chest or breast surgery, certain drugs (protease inhibitors, fluoroquinolones) and dyslipidaemia.^1,7

What happens in the joint capsule, and how does the condition evolve naturally?

The pathophysiology is a complex process combining synovial inflammation and capsular fibrosis. The early phase is marked by a hypervascular synovitis, with angiogenesis and an inflammatory infiltrate rich in mast cells, M2 macrophages and T lymphocytes.^1,8 The pro-inflammatory cytokines (IL-6, TNF-alpha) activate fibroblasts, which transform into myofibroblasts. These produce a disordered extracellular matrix rich in type I and III collagen, under the control of TGF-beta, the main mediator of capsular fibrosis.^1,9

This cascade leads to characteristic thickening at the coracohumeral ligament, the axillary recess and the rotator interval, progressively restricting joint volume.^1,10 In patients with diabetes, advanced glycation end-products (AGEs) interact with collagen and amplify capsular stiffness, which partly explains the more severe forms seen in this population.¹¹

The natural history is traditionally described as self-limiting, but that idea needs qualifying. The Hand 2008 study (223 patients, 269 shoulders, mean follow-up 4.4 years) found that 59 % of patients had normal or near-normal shoulders and 41 % reported persistent symptoms; these symptoms were mild in 94 % of cases and severe, with pain and functional loss, in 6 %, the outlook being worse the more severe the initial presentation.¹³ The retrospective series of Vastamäki 2012 (83 patients, follow-up 2 to 27 years) is more favourable: in untreated patients, range of motion returned to the level of the unaffected side in 94 % of cases, and 51 % were entirely pain-free.¹² The Wong 2017 meta-analysis confirms these data and identifies the adverse prognostic factors: diabetes, initial severity, depression and involvement of the dominant arm.¹⁴

Diagnosing frozen shoulder is essentially clinical. It rests on identifying a specific pattern of restricted glenohumeral movement, combined with actively excluding other conditions that can mimic the picture.^1,2 The official clinical practice guidelines of the American Physical Therapy Association (Kelley 2013 JOSPT, evidence level A) remain the international reference.³

Which questions should you ask to understand the patient and their history?

The history points strongly towards the diagnosis from the first consultation. The cardinal elements to look for are:

• Chronology : typically an insidious onset, with no triggering trauma, worsening gradually over several weeks to months.^1,3
• Pain profile : diffuse shoulder pain, often referred to the deltoid, worse at night and on rapid movement. Night pain is very frequent in the early phase but should not cause concern in the absence of other red flags.³
• Progressive functional limitation : difficulty doing one's hair, fastening a bra, reaching a back pocket or an object overhead.⁴
• Comorbidities to look for actively : diabetes mellitus (HbA1c, duration), thyroid dysfunction, Dupuytren's disease, a history of immobilisation, chest or breast surgery.^5,6
• Medication review : statins, protease inhibitors, fluoroquinolones and aromatase inhibitors have all been described.⁵

Which clinical tests should you perform, and which other conditions should you rule out?

The physical examination aims to confirm the symmetrical loss of both active AND passive range at the glenohumeral joint, which is the cardinal sign of frozen shoulder.^3,7

The Walmsley 2009 Delphi consensus (70 experts, three rounds) describes the early stage of primary frozen shoulder: marked night pain, pain with rapid or unguarded movement, discomfort lying on the affected shoulder, pain easily aggravated by movement, a global loss of active and passive range with end-range pain in all directions, and onset after the age of 35. The authors stress that these signs have limited ability to distinguish frozen shoulder from other shoulder conditions.⁷ Common operational definitions add:

• A loss of passive range of at least 25 % in at least 2 planes of movement compared with the healthy side.
• A loss of passive external rotation greater than 50 % (or less than 30 degrees in absolute terms) with the elbow at the side.
• A symmetrical limitation of active and passive movement (unlike cuff conditions).

The differential diagnosis is fundamental. The main conditions to rule out are:

• Rotator cuff conditions : tendinopathy, partial or massive tears. Passive movement is generally preserved, unlike in frozen shoulder. Jobe test, lift-off, bear hug, a painful arc in abduction (60-120 degrees).¹
• Glenohumeral osteoarthritis : similar stiffness, but the radiograph reveals joint space narrowing and osteophytes. The population is often older (> 60 years).²
• Calcific tendinopathy : very acute pain, calcification visible on the radiograph. An intense resorptive phase is possible.⁸
• Parsonage-Turner syndrome (neuralgic amyotrophy): disproportionate motor weakness, early muscle wasting, an abnormal EMG.⁹
• Cervical radicular conditions : neurological signs (paraesthesia, motor deficit in a root distribution), a positive Spurling test.¹
• Rare causes still worth considering : primary or metastatic tumour of the shoulder, septic arthritis, polymyalgia rheumatica (in older patients).¹⁰
• Missed posterior dislocation : after direct trauma, an epileptic seizure, an electric shock or electroconvulsive therapy. More than half of these dislocations go unnoticed at the first examination and are often mistaken for frozen shoulder. Blocked external rotation in this context calls for medical referral for imaging.¹³

Should patients be classified into stages, and what are the benefits?

Clinical staging by Reeves's phases (freezing, frozen, thawing), updated by Hannafin & Chiaia 2000 and then Millar 2022, remains useful for guiding treatment and managing the patient's expectations, even though this is a continuum rather than a set of discrete stages.^1,11

Frozen shoulder has a natural tendency to improve over 1 to 3 years, but its functional impact justifies active management. The 2020-2024 recommendations converge on a stepped multimodal strategy, matched to the phase of the condition and to tissue irritability.^1,2

Where do you start? What is the recommended hierarchy of interventions?

The initial approach rests on the triad of education + pain management + physiotherapy. The Challoumas 2020 network meta-analysis (JAMA Network Open, 65 randomised studies, 4,097 participants) compares the non-surgical treatments.³ It associates intra-articular corticosteroid injection, in frozen shoulder of less than one year's duration, with better short-term outcomes than the other non-surgical treatments, with a benefit that may last up to six months; this treatment should be accompanied by a home exercise programme. It could reach no conclusion on surgery, because the included trials did not compare the same interventions.³

A larger network meta-analysis (Meng 2026, BMJ, 189 RCTs, 12,988 participants) concludes, with moderate certainty, that extracorporeal shockwave therapy probably reduces short-term pain compared with usual care and that manipulation under anaesthetic, with or without injection, probably improves short-term flexion; most other estimates are of low or very low certainty, most trials are at high risk of bias, and comparative safety remains highly uncertain.¹⁸

The pragmatic randomised trial UK FROST (Rangan 2020 Lancet, 503 patients, 3 arms) is a major turning point in the literature: it compares early structured physiotherapy (with or without intra-articular corticosteroids), manipulation under anaesthetic and arthroscopic capsular release. At 12 months, no clinically significant difference is seen on the Oxford Shoulder Score (OSS). Physiotherapy is nonetheless the most cost-effective, and so remains the first-line intervention in primary and secondary care.⁴

What is the place of exercise and of manual therapy?

Supervised exercise plus a home exercise programme (HEP) is the cornerstone of conservative treatment. The Cochrane review Page 2014 (manual therapy + exercise, 32 RCTs) shows modest but consistent benefits on pain and function, particularly when exercise and manual therapy are combined.⁵

The principles of application are:^6,7

• Painful phase (freezing) : gentle pendulum exercises, active-assisted mobilisation within pain-free range. Avoid forced stretching, which can reactivate the inflammation.
• Stiff phase (frozen) : low-load, long-duration stretching (Russell 2014), end-range joint mobilisation (Maitland grade III-IV), a home exercise programme several times a day.
• Thawing phase : progressive strengthening of the cuff and the scapular stabilisers, a return to functional activities, prevention of contralateral recurrence.

The systematic review by Noten 2016 (APMR, 12 RCTs, 810 patients) concludes that mobilisation has beneficial effects and that the Maitland technique and combined mobilisations seem the ones to favour, on still preliminary evidence; in the only trial that assessed it, high-intensity mobilisation improved the Constant score less than no treatment.⁸ In a multiple-treatment trial of 28 patients, Yang 2007 found end-range mobilisation and mobilisation with movement (Mulligan) more effective than mid-range mobilisation for mobility and function.⁹

Injections, hydrodilatation, surgery: when should you escalate?

The intra-articular corticosteroid injections are an option, particularly in phase 1 (painful): the 2025 Korean clinical practice guideline suggests considering them, alone or combined with physiotherapy (conditional recommendation, low level of evidence).¹⁷ The Sun 2017 meta-analysis (AJSM, 8 RCTs, 416 patients) shows, compared with no injection or a sham injection, a significant reduction in pain at 4-6 weeks and at 12-16 weeks.¹⁰ The clinical superiority of ultrasound guidance has not been demonstrated, its use depending on the operator's expertise, and the dose of triamcinolone (20 or 40 mg) is a matter of clinical judgement.¹⁷

Arthrographic hydrodilatation (distension with a large volume of saline plus corticosteroid plus anaesthetic under ultrasound or fluoroscopic guidance) is an option for patients who do not respond to 3-6 months of conservative treatment, particularly patients with diabetes. The retrospective study Dimitri-Pinheiro 2022 (PMC9465918) shows reduced but real efficacy in patients with diabetes compared with those without.¹¹

Any surgery (manipulation under anaesthetic or arthroscopic release) remains reserved for failures of maximal conservative management (beyond 6-9 months). UK FROST shows that none of the three arms (physiotherapy with injection, manipulation under anaesthetic, arthroscopic release) is clinically superior at 12 months, the differences remaining below the thresholds set in advance; arthroscopic release carries more serious adverse events (eight, against two after manipulation), which justifies favouring conservative management for its lower invasiveness.⁴

Beyond the physical: how do you educate the patient and address psychological factors?

Therapeutic education is the indispensable first step. It aims to:^2,12

• Explain the natural history (3 phases, a typical duration of 1-3 years, the possibility of minor residual impairment).
• Reassure the patient that the condition is not serious (in the absence of red flags).
• Counter the kinesiophobia and catastrophising that worsen disability (Chester 2018 BJSM).
• Promote self-management (the home exercise programme, energy management, adapting ADLs).

The scoping review Coronado 2020 (Pain Rep) highlights the value of psychologically informed physiotherapy for patients with yellow flags: catastrophising, fear of movement, low self-efficacy.¹³

Frozen shoulder in the patient with diabetes is a distinct clinical entity in its frequency, its severity and its resistance to standard treatment. Recognising that specificity is essential for adapting management and setting realistic expectations.^1,2

What are the biological mechanisms linking diabetes and frozen shoulder?

The Zreik 2016 meta-analysis (MLTJ, 12 studies) establishes that the prevalence of frozen shoulder reaches 13.4 % in people with diabetes (against 2-5 % in the general population) and, conversely, that about 30 % of patients with frozen shoulder have diabetes (95 % CI 24-37).³ The narrative review Chen 2024 (J Clin Med, PMID 39407755) synthesises the biological mechanisms identified to date:¹

• Advanced glycation end-products (AGEs) : chronic hyperglycaemia leads to the formation of AGEs that bind to the collagen of the joint capsule. These AGE cross-links stiffen the extracellular matrix and resist normal degradation mechanisms.
• Overexpression of TGF-beta and amplification of the profibrotic signalling pathway in people with diabetes, comparable to what is seen in Dupuytren's disease (another fibrosis associated with diabetes).
• Microangiopathy : impaired periarticular microcirculation, with tissue hypoxia favouring fibrosis.
• Chronic low-grade inflammation linked to diabetes (raised CRP, IL-6), which prolongs the synovitic phase.
• Partial resistance to corticosteroids : a poorer response to intra-articular corticosteroids than in people without diabetes (Dimitri-Pinheiro 2022).

How do you stratify risk and adapt management in the patient with diabetes?

Clinical stratification helps personalise the prognosis and the treatment plan. Several predictors of severity and resistance are documented:^1,4,5

• Duration of diabetes : the risk rises beyond 10 years.
• HbA1c > 7.5 % : poor glycaemic control is associated with slower recovery and a higher failure rate.
• Type 1 diabetes : its own role, like that of prolonged insulin therapy, remains debated (Chen 2024).
• The presence of other microvascular complications : retinopathy, nephropathy, neuropathy.
• A bilateral or recurrent form.

Ultrasound-guided hydrodilatation is an option of choice in the patient with diabetes and resistant frozen shoulder. The Mezian StatPearls review and the Dimitri-Pinheiro 2022 study (PMC9465918) show that it allows rehabilitation to be restarted in about 50-70 % of patients who stall despite 3-6 months of well-conducted PT, with a favourable safety profile.⁴ Important: tell the patient with diabetes about the transient risk of hyperglycaemia after a corticosteroid injection (peak effect at 24-72 h) and coordinate with their general practitioner about adjusting insulin if needed.¹

Frozen shoulder is often presented as a self-limiting condition, which can lead to underestimating the importance of structured management. Yet, as Hand 2008 showed, about 40 % of patients still have symptoms in the long term, most often mild.² A well-conducted secondary prevention programme is therefore essential.

How do you make the patient an active participant in their recovery through self-management?

Self-management (self-management) guided by the physiotherapist is the pillar of durable recovery. It rests on three pillars:^3,4

1. Education about the natural history : explain the 3 phases, the typical duration (1-3 years), how rare recurrence on the same side is (under 5 %) but the risk of contralateral involvement (6-17 % according to Wong 2017, above all in people with diabetes).
2. A home exercise programme (HEP) :
   • Done several times a day (3-5 sessions of 10-15 minutes).
   • Controlled progression, respecting the pain threshold (the 24 h rule: if pain increases the next day, reduce the intensity).
   • Adapt to the phase: pendulum exercises and gentle mobilisation in phase 1; low-load, long-duration stretching in phase 2; progressive strengthening in phase 3.
3. Adaptive management of daily activities :
   • Adapt the provocative movements without giving up movement altogether.
   • Support the painful arm at night (a pillow under the elbow).
   • Apply heat before the exercises, ice afterwards if there is post-exertional pain.

When and how should a safe return to sport and to activity be planned?

The return to sport (RTS) or to demanding occupational activity (lifting, overhead work) should be based on objective functional criteria, not on a fixed timescale. The consensus criteria are:^5,6

• Range : passive and active ROM above 90 % of the healthy side, particularly ER and elevation.
• Strength : isometric cuff strength (Jobe, ER in abduction, lift-off) above 90 % of the other side.
• Pain : VAS < 2/10 for the movements specific to the sport or the job.
• Neuromuscular control : no scapular dyskinesis, proprioceptive control of complex movements.
• Confidence and self-efficacy : assessed by questionnaire (the TSK for fear of movement, for example).

The working environment should also be considered. For demanding jobs (manual handling, care work, ceiling work, painting, overhead sports), a graded return with workplace adjustments is often needed for 2-3 months after the thawing phase ends.⁵

Analysis of a “classic” case: from assessment to resolution.

The case report Phansopkar 2023 (Cureus, PMC10387580) describes a 57-year-old man with a painful, stiff right shoulder for more than six weeks, with lateral rotation of 33° against 80° on the unaffected side. Over three weeks, at five sessions a week, he received moist heat, Codman's pendulum exercises, scapular setting exercises, interferential therapy and the Spencer technique (rotation of the humerus around the elbow used as a pivot, with the arm abducted, then with traction). Pain at rest fell from 7/10 to 0/10 and the SPADI from 72 % to 30 % at six months.¹

The diagnostic challenge: when frozen shoulder mimics another condition.

The case published by Ohta 2017 (J Rural Med, PMID 29255532) is that of a 28-year-old woman seen for acute pain in both shoulders, with, from the first examination, wasting of the right deltoid, weakness of every movement of the right shoulder and sensory loss over its lateral aspect. MRI and electromyography confirmed bilateral neuralgic amyotrophy (Parsonage-Turner syndrome). The case is a reminder of the signs that should rule out a diagnosis of frozen shoulder: acute onset, muscle wasting and motor deficit.²

The warning signs that should prompt the diagnosis of frozen shoulder to be reconsidered:^2,3

• Very intense, abrupt initial pain, sometimes preceded by a viral illness.
• Marked motor weakness, greater than the loss of passive range.
• Early, visible muscle wasting (deltoid, supraspinatus/infraspinatus, serratus anterior).
• Failure of well-conducted PT after 6-8 weeks.
• A territorial distribution of the weakness (suprascapular nerve, long thoracic nerve).

Similarly, calcific tendinopathy can keep a shoulder very painful; a plain shoulder radiograph (AP and lateral) will show it.

A complex case: resistant frozen shoulder in a patient with diabetes.

The retrospective study Dimitri-Pinheiro 2022 (PMC9465918) is an excellent illustration of managing resistant frozen shoulder in the patient with diabetes. A series of patients with diabetes and frozen shoulder refractory to 3-6 months of conventional PT received ultrasound-guided hydrodilatation (injection of 30-40 mL of saline plus corticosteroid plus lidocaine under posterior ultrasound guidance).⁵

The results show:⁵

• An immediate mobility gain (ER, ABD) of the order of 20-30 degrees.
• Progressive improvement over 6 months with PT continuing.
• A response that is smaller than in people without diabetes, but clinically significant.
• A favourable safety profile, provided insulin is adjusted around the procedure.

This kind of case illustrates the treatment escalation strategy : start with PT + HEP + intra-articular corticosteroids in phase 1, then consider hydrodilatation if progress plateaus at 3-6 months, and keep MUA or arthroscopy as a last resort, following UK FROST.⁶

When and to which other health professionals should you refer?

Collaborative referral is an essential dimension of practice. The situations calling for medical referral are:

• The presence of red flags (reasoning transposed from the IFOMPT framework of Finucane 2020, which was built for serious spinal pathology): non-mechanical night pain, weight loss, persistent fever, a history of cancer, progressive neurological involvement, marked inflammatory signs.¹
• Stalled progress at 3-6 months despite well-conducted PT: refer to an orthopaedic surgeon to consider intra-articular corticosteroids, hydrodilatation or surgery.²
• A suspected alternative diagnosis (Parsonage-Turner, refractory calcific tendinopathy, early inflammatory arthritis): rheumatologist or neurologist.
• A patient with poorly controlled diabetes (HbA1c > 8 %): diabetologist, for glycaemic optimisation.
• Significant psychological distress (catastrophising, depression): psychologist or psychotherapist (CBT).^3,4

Structured interprofessional collaboration is a determinant of care quality. The key principles include: goals shared with the patient, structured communication between clinicians (a formal physiotherapy report, a letter to the general practitioner), and mutual respect for scopes of practice.⁵

How do you measure outcomes and overcome barriers to implementation?

Standardised outcome measurement is indispensable for making progress objective and adapting management. For frozen shoulder, the PROMs validated in French and most widely used are:⁶

The VAS complements the PROMs: it allows quick day-to-day tracking of pain. Physical performance tests (goniometric ROM measurement, isometric strength testing with a dynamometer) remain indispensable and complete the PROMs for an overall picture.⁶

Overcoming the barriers to implementing the evidence

The Scurlock-Evans 2014 meta-analysis (Physiotherapy, PMID 24780633) identifies the main barriers to implementing evidence-based practice in physiotherapy:⁷

• Lack of time in clinical practice.
• Difficulty accessing and critically appraising the literature.
• Lack of organisational support.
• The gap between the reality of complex patients and the often homogeneous populations of RCTs.

The most effective improvement strategies are:^8,9

• Structured audit and feedback (Ivers 2012 Cochrane): comparing one's practice with evidence-based indicators, followed by constructive feedback.
• Identifying local leaders (Flodgren 2019 Cochrane) who promote and support the change.
• Interactive meetings and workshops as continuing education.
• Decision-support tools built into the practice software.